The entire paper rests on discussions at a session featuring top PSE-ML researchers. This creates a massive selection bias.
No Skeptics? Where were the experts who question if ML is the real bottleneck, and not, say, our ability to solve complex non-convex problems with traditional methods?
Echo Chamber? With a pre-selected, pro-ML group, the challenges identified simply validate their own research agendas.
Black Box Discussions: The paper gives zero methodology for how consensus was reached. Were dissenting views even presented or recorded?
Without a transparent, critical process, this isn’t a field-wide roadmap; it’s a summary of an insider meeting, potentially misdirecting the community’s priorities.